Impact of widespread LLM (AI) availability on student exam performance

This study examines whether the widespread availability of artificial intelligence (AI) and large language model (LLM) tools is associated with changes in student academic performance in an online Business Statistics course. Using a dataset of 757 students taught by the same instructor between 2018 and 2026, the analysis compares final cumulative exam scores from pre-AI (2018–2022) and post-AI (2023–2026) cohorts. An independent sample t-test, Mann-Whitney U test, and regression analysis are employed to assess differences in performance and ensure robustness of results. The findings consistently indicate a statistically significant increase in exam scores in the post-AI period. The nonparametric test confirms a significant upward shift in score distributions, while regression results provide a consistent estimate of the performance gain associated with AI availability. Overall, the results suggest that student performance improved following the introduction of AI tools, although the magnitude of the effect is modest. These findings highlight the need to reconsider assessment practices and academic integrity policies in the context of rapidly evolving educational technologies.